The Epoch-State Filter, Revisited

Abstract

The epoch-state filter is a recursive sequential estimator that maintains its state estimate and associated covariance at a fixed time known as the epoch, avoiding the need to propagate the covariance between measurements. Instead, the filter maps measurement innovations from the measurement time to the epoch time, in a fashion similar to a batch estimator. When process noise is present, the epoch time must follow all but at most the penultimate measurement. Revisiting the topic, this paper contributes several variations, including a factorized form, a factorized information form, and sigma-point versions, and provides a number of clarifying illustrative examples.